In the late 1980’s, psychologist Ole Ivar Lovaas argued that the optimal amount for a behavior therapy was around 40 hours a week. This type of intensive therapy, termed  applied behavior analysis (ABA), led to a “normal” functioning 9 of 19 children with autism. Emphasizing that Just 10 hours of the therapy produced small benefits. Small controlled trials of Lovaas’ therapy, however, did not fully replicate his findings on dose.

The form of ABA Lovaas tested has become the most widely used autism therapy — despite about its approach and findings — and 40 hours became a treatment goal for many families. ABA agencies typically market services at that intensity. And this year, that goal was formalized by the Behavior Analyst Certification Board, which recommended 30 to 40 hours of treatment per week for autistic children who need help in several areas, such as cognition, communication or social functioning. They recommend fewer hours for children with fewer needs.)

Receiving 40 hours of weekly therapy is essentially a full-time job for a 3-year-old.

Still, people tend to believe that, regardless of the treatment, more is always better. But is it?

In an article published earlier this year, a team of researchers rightfully questions these specific requirements for autistic children. They note, correctly, that research does not point to an optimal frequency or duration — the ‘dose,’ if you will — for ABA. In fact, a universal recommendation is unlikely to be appropriate given the heterogeneity of autism.

We need to better understand what dose is effective for which types of behavioral therapy, and for which children. Our goal should be to develop potent therapies that can be delivered within a reasonable time frame and that can be personalized for autistic children with distinct needs.

Defining success:

It is possible that all early interventions are more effective at high doses than at ones. Or it may be that the intensive form of ABA Lovaas tested, called  discrete trial training, requires more hours than other therapies do to provide benefit. We simply do not know which alternative is true, given the lack of rigorous studies comparing interventions.

In one common study design, researchers compare a new treatment with an established therapy. But which established therapy controls get may differ among the participants. And children may get a lot of therapy or a little: Most studies do not control for dose.

If the experimental group does significantly better than controls do, it is easy to conclude that the new therapy is optimal. But if the children receiving the new therapy get more treatment than the other children do, the higher dose may have contributed to the difference — and it is impossible to tell how much.

Other studies attempt to quantify may conflate scheduled therapy hours with received therapy hours. Children may not receive the hours they are scheduled for due to program changes, absences, illnesses or other constraints. So these types of quantification remain crude, and quality of the therapies.

There are ways to control for dose so that interpreting results is more straightforward.

Modular approach:

Given the great differences among children with autism, it is unlikely that all children require the same intervention equally, or even the same amount of time a similar intervention.

Another approach found within long-term studies indicated that about half of all children with autism make noteworthy progress with early intensive interventions that require at least 20 hours per week. Other children make slower progress, with about 30 percent remaining minimally verbal at school age despite intensive treatment.

As we move toward greater customization of therapy, applying a modular approach whereby some children receive one type of treatment (i.e. training) to address some treatment goals and such as developmental method for other goals.

Some children may need high intensity at first and then lower over time as they improve. However, in practice a more economical approach would be to start all children at low intensity, assess each child’s response at a prespecified time and step up the intensity if the response is slow.

New research methodologies allow us to test the responses to these models. Ultimately, these models will help us provide effective early therapy to all children who need it, including those with few resources or who live far away from specialists.


  1. O.I. J. Consult. Clin. Psychol. 55, 3-9 (1987) PubMed
  2. . Am. J. . Retard. 105, 269-285 (2000) PubMed
  3. Sallows G.O. T.D. Graupner Am. J. . Retard. 110, 417-438 (2005) PubMed
  4. M.  Autism 23, 1075-1078 (2019) PubMed
  5. C. T. Smith Clin. Psychol- Sci. . 23, 260-264 (2016) Full text
  6. C.  J. Child Psychol. Psychiatry 47, 611-620 (2006) PubMed
  7. C.  J. Consult. Clin. Psychol. 83, 554-563 (2015) PubMed
  8. S.  Am. J. Intellect. Dev. 115, 381-405 (2010) PubMed
  9. Tager‐Flusberg H. C.  Autism Res. 6, 468-478 (2013) PubMed
  10. Almirall D. A. Chronis-Tuscano J. Clin. Child Adolesc. Psychol. 45, 383-395 (2016) PubMed